Research Interests

Prof Richtarik develops new algorithms for solving optimization and machine learning problem involving very large quantities of data. His recent work is focused on randomized algorithms of various flavors, including randomized coordinate descent, stochastic gradient descent, randomized subspace descent and randomized Newton and quasi-Newton methods. Parallel and distributed variants are of particular importance.


Selected Publications

  • Federated optimization: distributed machine learning for on-device intelligence
    J. Konecny, H. B. McMahan, D. Ramage and P. Richtarik
    Android, (2016)
  • Randomized iterative methods for linear systems
    R. M. Gower, P. Richtarik
    SIAM Journal on Matrix Analysis and Applications 36(4), 1660-1690, (2015)
  • Mini-batch semi-stochastic gradient descent in the proximal setting
    J. Konecny, J. Liu, P. Richtarik and M. Takac
    IEEE Journal of Selected Topics in Signal Processing 10(2), 242-255, (2016)
  • Stochastic dual coordinate ascent with adaptive probabilities
    D. Csiba, Z. Qu and P. Richtarik
    In Proceedings of The 32nd International Conference on Machine Learning, 674-683..., (2015)
  • Adding vs. averaging in distributed primal-dual optimization
    C. Ma, V. Smith, M. Jaggi, M. I. Jordan, P. Richtarik and M. Takac
    In Proceedings of The 32nd International Conference on Machine Learning, 1973-19..., ( 2015)
  • Quartz: Randomized dual coordinate ascent with arbitrary sampling
    Z. Qu, P. Richtarik and T. Zhang
    In Advances in Neural Information Processing Systems 28, 865-873, (2015)
  • Accelerated, Parallel and PROXimal coordinate descent
    O. Fercoq and P. Richtarik
    SIAM Journal on Optimization 25(4), 1997-2023, (2015)
  • Parallel coordinate descent methods for big data optimization
    P. Richtarik and M. Takac
    Mathematical Programming 156(1), 433-484, (2016)
  • Iteration complexity of randomized block-coordinate descent methods for minimizing a composite function
    P. Richtarik and M. Takac
    Mathematical Programming 144(2), 1-38, (2014)
  • Generalized power method for sparse principal component analysis
    M. Journee, Yu. Nesterov, P. Richtarik and R. Sepulchre
    Journal of Machine Learning Research 11, 517553, (2010)

Education Profile

PhD, Operations Research, Cornell University, 2002–2007, advisor: Mike Todd
MS, Operations Research, Cornell University, 2006
Mgr, Mathematics, Comenius University, Faculty of Math, Physics & Informatics, 2001 (Summa Cum Laude)
Bc, Management, Comenius University, Faculty of Management, 2000 (Summa Cum Laude)
Bc, Mathematics, Comenius University, Faculty of Math, Physics & Informatics, 2000 (Summa Cum Laude)


Professional Profile

2017-present: Associate Professor, Computer Science, Mathematics, KAUST, Saudi Arabia

2016-present: Associate Professor, School of Mathematics, University of Edinburgh, United Kingdom

2015-present: Faculty Fellow, The Alan Turing Institute, London, United Kingdom

2013: Visiting Assistant Professor, University of California, Berkeley, USA

2009-2016: Assistant Professor, School of Mathematics, University of Edinburgh, United Kingdom

2007-2009: Postdoctoral Fellow, CORE & INMA, Louvain-la-Neuve, Belgium


Scientific and Professional Memberships

Society for Industrial and Applied Mathematics (SIAM)

Mathematical Optimization Society (MOS)

Edinburgh Mathematical Society (EMS)

Isaac Newton Institute for Mathematical Sciences (INIMS)

Institute for Operations Research and Management Science (INFORMS)

Foundations of Computational Mathematics (FoCM)


Awards

2016, SIAM SIGEST Award

2016, EUSA Best Research Supervisor Award (2nd place), University of Edinburgh

2016, EPSRC Fellowship in Mathematical Sciences

2016, Plenary, 41st Woudschoten Conference, The Netherlands

2016, Einstein Center Mathematical Colloquium

2014, Plenary, 46th Conference of Slovak Mathematicians

2013, Plenary, NIPS Workshop on Optimization in Machine Learning


KAUST Affiliations

Visual Computing Center (VCC)

Extreme Computing Research Center (ECRC)

Computer, Electrical and Mathematical Science and Engineering (CEMSE)